Using Docker Compose v3, it's possible to define a Docker deployment along with production details. This provides a central location for managing your application deployments that can be deployed onto a Swarm Mode cluster.

A Docker Compose file has been created that defines deploying a Redis server with a web front end.

View the file using cat docker-compose.yml

The file has been extended to utilize Swarm deployment options.

The first configuration option uses depends_on. This states that Redis must be deployed before the web and allows us to control the order of services being started.

The next configuration options define how the application should be deployed using the new deploy options.

Firstly, mode: replicated and replicas: 2 determine how many replicas of the service should be started.

Secondly, resources are define. The limits are hard limits that the application cannot exceed, the reservations is a guide to Docker Swarm to indicate the resources the applications requires.

Third, restart_policy indicates what should happen if the process crashes.

Fourth, update_config defines how updates should be applied and rolled out.

Finally, placement allows us to add constraints to determine where the service should be deployed.

More details can be found at https://docs.docker.com/compose/compose-file/#deploy

The Docker Compose file is referred to as a Docker Compose Stack. Stacks can be deployed to Swarm using the CLI.

Task

The docker stack command is used to deploy a Docker Compose Stack via Swarm. In this case, it will prefix the services with myapp.

docker stack deploy --compose-file docker-compose.yml myapp

Once deployed it's possible to use the CLI to inspect the state.

The command docker stack ls lists all stacks deployed.

Details of the internal services can be discovered via docker stack services myapp

Notice that the command indicates the Desired / Running state for the service. If the service cannot be deployed then this will be different.

The details of each service container can be identified using docker stack ps myapp

All of this information can still be discovered using docker ps